Hugging Face's logo

fxmarty
/
donotdelete2

Question Answering
TensorBoard
distilbert
Model card Files Metrics Community
fxmarty commited on
Commit
927a97d
路
1 Parent(s): 75a98e2

add experience

Browse files
Files changed (7) hide show
  1. README.md +51 -0
  2. runs.json +456 -0
  3. tensorboard/1652945902.9702048/events.out.tfevents.1652945902.huggingface.8.1 +3 -0
  4. tensorboard/1652945902.971785/events.out.tfevents.1652945902.huggingface.8.2 +3 -0
  5. tensorboard/1652945902.9734063/events.out.tfevents.1652945902.huggingface.8.3 +3 -0
  6. tensorboard/1652945902.9749742/events.out.tfevents.1652945902.huggingface.8.4 +3 -0
  7. tensorboard/events.out.tfevents.1652945902.huggingface.8.0 +3 -0
README.md ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pipeline_tag: question-answering
3
+ datasets:
4
+ - squad
5
+ metrics:
6
+ - exact_match
7
+ - f1
8
+ tags:
9
+ - distilbert
10
+ ---
11
+
12
+ **task**: `question-answering`
13
+
14
+ Fixed parameters:
15
+ * **model_name_or_path**: `distilbert-base-uncased-distilled-squad`
16
+ * **dataset**:
17
+ * **path**: `squad`
18
+ * **name**: `None`
19
+ * **calibration_split**: `train`
20
+ * **eval_split**: `validation`
21
+ * **data_keys**: `{'question': 'question', 'context': 'context'}`
22
+ * **ref_keys**: `['answers']`
23
+ * **max_seq_length**: `128`
24
+ * **node_exclusion**: `[]`
25
+ * **per_channel**: `False`
26
+ * **calibration**:
27
+ * **method**: `minmax`
28
+ * **num_calibration_samples**: `20`
29
+ * **calibration_histogram_percentile**: `None`
30
+ * **calibration_moving_average**: `None`
31
+ * **calibration_moving_average_constant**: `None`
32
+ * **framework**: `onnxruntime`
33
+ * **framework_args**:
34
+ * **opset**: `15`
35
+ * **optimization_level**: `1`
36
+ * **aware_training**: `False`
37
+
38
+ Benchmarked parameters:
39
+ * **quantization_approach**: `dynamic`, `static`
40
+ * **operators_to_quantize**: `['Add', 'MatMul']`, `['Add']`
41
+
42
+ ## Evaluation
43
+ Below, time metrics for
44
+ * Batch size: 8
45
+ * Input length: 128
46
+ | quantization_approach | operators_to_quantize | | latency_mean (original, ms) | latency_mean (optimized, ms) | | throughput (original, /s) | throughput (optimized, /s) | | exact_match (original) | exact_match (optimized) | | f1 (original) | f1 (optimized) |
47
+ | :-------------------: | :-------------------: | :-: | :-------------------------: | :--------------------------: | :-: | :-----------------------: | :------------------------: | :-: | :--------------------: | :---------------------: | :-: | :-----------: | :------------: |
48
+ | `dynamic` | `['Add', 'MatMul']` | \| | 247.65 | 75.80 | \| | 4.50 | 13.50 | \| | 90.000 | 85.000 | \| | 90.000 | 86.667 |
49
+ | `dynamic` | `['Add']` | \| | 234.61 | 191.78 | \| | 4.50 | 5.50 | \| | 90.000 | 90.000 | \| | 90.000 | 90.000 |
50
+ | `static` | `['Add', 'MatMul']` | \| | 238.04 | 131.81 | \| | 4.50 | 8.00 | \| | 90.000 | 50.000 | \| | 90.000 | 59.362 |
51
+ | `static` | `['Add']` | \| | 241.84 | 257.85 | \| | 4.50 | 4.00 | \| | 90.000 | 70.000 | \| | 90.000 | 76.333 |
runs.json ADDED
@@ -0,0 +1,456 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "model_name_or_path": "distilbert-base-uncased-distilled-squad",
4
+ "task": "question-answering",
5
+ "dataset": {
6
+ "path": "squad",
7
+ "name": null,
8
+ "calibration_split": "train",
9
+ "eval_split": "validation",
10
+ "data_keys": {
11
+ "question": "question",
12
+ "context": "context"
13
+ },
14
+ "ref_keys": [
15
+ "answers"
16
+ ],
17
+ "max_seq_length": 128
18
+ },
19
+ "quantization_approach": "dynamic",
20
+ "operators_to_quantize": [
21
+ "Add",
22
+ "MatMul"
23
+ ],
24
+ "node_exclusion": [],
25
+ "aware_training": false,
26
+ "per_channel": false,
27
+ "calibration": {
28
+ "method": "minmax",
29
+ "num_calibration_samples": 20,
30
+ "calibration_histogram_percentile": null,
31
+ "calibration_moving_average": null,
32
+ "calibration_moving_average_constant": null
33
+ },
34
+ "framework": "onnxruntime",
35
+ "framework_args": {
36
+ "opset": 15,
37
+ "optimization_level": 1
38
+ },
39
+ "hardware": "Architecture: x86_64\nCPU op-mode(s): 32-bit, 64-bit\nByte Order: Little Endian\nAddress sizes: 46 bits physical, 48 bits virtual\nCPU(s): 20\nOn-line CPU(s) list: 0-19\nThread(s) per core: 1\nCore(s) per socket: 14\nSocket(s): 1\nNUMA node(s): 1\nVendor ID: GenuineIntel\nCPU family: 6\nModel: 154\nModel name: 12th Gen Intel(R) Core(TM) i7-1280P\nStepping: 3\nCPU MHz: 1195.476\nCPU max MHz: 4800.0000\nCPU min MHz: 400.0000\nBogoMIPS: 3993.60\nVirtualization: VT-x\nL1d cache: 336 KiB\nL1i cache: 224 KiB\nL2 cache: 8.8 MiB\nNUMA node0 CPU(s): 0-19\nVulnerability Itlb multihit: Not affected\nVulnerability L1tf: Not affected\nVulnerability Mds: Not affected\nVulnerability Meltdown: Not affected\nVulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp\nVulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2: Mitigation; Enhanced IBRS, IBPB conditional, RSB filling\nVulnerability Srbds: Not affected\nVulnerability Tsx async abort: Not affected\nFlags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l2 invpcid_single cdp_l2 ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdt_a rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req umip pku ospke waitpkg gfni vaes vpclmulqdq tme rdpid movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr flush_l1d arch_capabilities\n",
40
+ "versions": {
41
+ "transformers": "4.19.2",
42
+ "optimum": "1.2.1.dev0",
43
+ "optimum_hash": "f5a741330bef093b6ffe0023455dff722e4515e8"
44
+ },
45
+ "evaluation": {
46
+ "time": [
47
+ {
48
+ "batch_size": 8,
49
+ "input_length": 128,
50
+ "baseline": {
51
+ "nb_forwards": 9,
52
+ "throughput": 4.5,
53
+ "latency_mean": 247.6471492222222,
54
+ "latency_std": 10.094966731742604,
55
+ "latency_50": 248.684487,
56
+ "latency_90": 258.2063756,
57
+ "latency_95": 261.8118248,
58
+ "latency_99": 264.69618416,
59
+ "latency_999": 265.345165016
60
+ },
61
+ "optimized": {
62
+ "nb_forwards": 27,
63
+ "throughput": 13.5,
64
+ "latency_mean": 75.79600677777778,
65
+ "latency_std": 10.523121143960658,
66
+ "latency_50": 72.788173,
67
+ "latency_90": 93.5972356,
68
+ "latency_95": 97.55360420000001,
69
+ "latency_99": 102.62172174,
70
+ "latency_999": 103.992470574
71
+ }
72
+ },
73
+ {
74
+ "batch_size": 4,
75
+ "input_length": 128,
76
+ "baseline": {
77
+ "nb_forwards": 16,
78
+ "throughput": 8.0,
79
+ "latency_mean": 125.356500125,
80
+ "latency_std": 7.659526175474619,
81
+ "latency_50": 120.1321565,
82
+ "latency_90": 136.778588,
83
+ "latency_95": 138.5083155,
84
+ "latency_99": 139.35583590000002,
85
+ "latency_999": 139.54652799000002
86
+ },
87
+ "optimized": {
88
+ "nb_forwards": 54,
89
+ "throughput": 27.0,
90
+ "latency_mean": 37.474536,
91
+ "latency_std": 2.0764437966332765,
92
+ "latency_50": 37.084555,
93
+ "latency_90": 40.0797993,
94
+ "latency_95": 41.22111285,
95
+ "latency_99": 42.984754949999996,
96
+ "latency_999": 43.687130295
97
+ }
98
+ }
99
+ ],
100
+ "others": {
101
+ "baseline": {
102
+ "exact_match": 90.0,
103
+ "f1": 90.0
104
+ },
105
+ "optimized": {
106
+ "exact_match": 85.0,
107
+ "f1": 86.66666666666667
108
+ }
109
+ }
110
+ },
111
+ "metrics": [
112
+ "squad"
113
+ ],
114
+ "model_type": "distilbert"
115
+ },
116
+ {
117
+ "model_name_or_path": "distilbert-base-uncased-distilled-squad",
118
+ "task": "question-answering",
119
+ "dataset": {
120
+ "path": "squad",
121
+ "name": null,
122
+ "calibration_split": "train",
123
+ "eval_split": "validation",
124
+ "data_keys": {
125
+ "question": "question",
126
+ "context": "context"
127
+ },
128
+ "ref_keys": [
129
+ "answers"
130
+ ],
131
+ "max_seq_length": 128
132
+ },
133
+ "quantization_approach": "static",
134
+ "operators_to_quantize": [
135
+ "Add",
136
+ "MatMul"
137
+ ],
138
+ "node_exclusion": [],
139
+ "aware_training": false,
140
+ "per_channel": false,
141
+ "calibration": {
142
+ "method": "minmax",
143
+ "num_calibration_samples": 20,
144
+ "calibration_histogram_percentile": null,
145
+ "calibration_moving_average": null,
146
+ "calibration_moving_average_constant": null
147
+ },
148
+ "framework": "onnxruntime",
149
+ "framework_args": {
150
+ "opset": 15,
151
+ "optimization_level": 1
152
+ },
153
+ "hardware": "Architecture: x86_64\nCPU op-mode(s): 32-bit, 64-bit\nByte Order: Little Endian\nAddress sizes: 46 bits physical, 48 bits virtual\nCPU(s): 20\nOn-line CPU(s) list: 0-19\nThread(s) per core: 1\nCore(s) per socket: 14\nSocket(s): 1\nNUMA node(s): 1\nVendor ID: GenuineIntel\nCPU family: 6\nModel: 154\nModel name: 12th Gen Intel(R) Core(TM) i7-1280P\nStepping: 3\nCPU MHz: 2000.000\nCPU max MHz: 4800.0000\nCPU min MHz: 400.0000\nBogoMIPS: 3993.60\nVirtualization: VT-x\nL1d cache: 336 KiB\nL1i cache: 224 KiB\nL2 cache: 8.8 MiB\nNUMA node0 CPU(s): 0-19\nVulnerability Itlb multihit: Not affected\nVulnerability L1tf: Not affected\nVulnerability Mds: Not affected\nVulnerability Meltdown: Not affected\nVulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp\nVulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2: Mitigation; Enhanced IBRS, IBPB conditional, RSB filling\nVulnerability Srbds: Not affected\nVulnerability Tsx async abort: Not affected\nFlags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l2 invpcid_single cdp_l2 ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdt_a rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req umip pku ospke waitpkg gfni vaes vpclmulqdq tme rdpid movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr flush_l1d arch_capabilities\n",
154
+ "versions": {
155
+ "transformers": "4.19.2",
156
+ "optimum": "1.2.1.dev0",
157
+ "optimum_hash": "f5a741330bef093b6ffe0023455dff722e4515e8"
158
+ },
159
+ "evaluation": {
160
+ "time": [
161
+ {
162
+ "batch_size": 4,
163
+ "input_length": 128,
164
+ "baseline": {
165
+ "nb_forwards": 16,
166
+ "throughput": 8.0,
167
+ "latency_mean": 129.7835274375,
168
+ "latency_std": 8.180046367856916,
169
+ "latency_50": 127.4055855,
170
+ "latency_90": 140.2466055,
171
+ "latency_95": 141.5487165,
172
+ "latency_99": 143.7097641,
173
+ "latency_999": 144.19599981
174
+ },
175
+ "optimized": {
176
+ "nb_forwards": 33,
177
+ "throughput": 16.5,
178
+ "latency_mean": 61.21133818181818,
179
+ "latency_std": 4.1161118662371425,
180
+ "latency_50": 60.193171,
181
+ "latency_90": 67.8919568,
182
+ "latency_95": 69.4033968,
183
+ "latency_99": 70.35858768000001,
184
+ "latency_999": 70.63236076800001
185
+ }
186
+ },
187
+ {
188
+ "batch_size": 8,
189
+ "input_length": 128,
190
+ "baseline": {
191
+ "nb_forwards": 9,
192
+ "throughput": 4.5,
193
+ "latency_mean": 238.0396341111111,
194
+ "latency_std": 10.880216114686617,
195
+ "latency_50": 234.615554,
196
+ "latency_90": 253.6232216,
197
+ "latency_95": 257.6177148,
198
+ "latency_99": 260.81330936,
199
+ "latency_999": 261.532318136
200
+ },
201
+ "optimized": {
202
+ "nb_forwards": 16,
203
+ "throughput": 8.0,
204
+ "latency_mean": 131.8145219375,
205
+ "latency_std": 13.714115627838893,
206
+ "latency_50": 124.4032245,
207
+ "latency_90": 152.588659,
208
+ "latency_95": 154.6390445,
209
+ "latency_99": 156.57490969999998,
210
+ "latency_999": 157.01047937
211
+ }
212
+ }
213
+ ],
214
+ "others": {
215
+ "baseline": {
216
+ "exact_match": 90.0,
217
+ "f1": 90.0
218
+ },
219
+ "optimized": {
220
+ "exact_match": 50.0,
221
+ "f1": 59.36204146730463
222
+ }
223
+ }
224
+ },
225
+ "metrics": [
226
+ "squad"
227
+ ],
228
+ "model_type": "distilbert"
229
+ },
230
+ {
231
+ "model_name_or_path": "distilbert-base-uncased-distilled-squad",
232
+ "task": "question-answering",
233
+ "dataset": {
234
+ "path": "squad",
235
+ "name": null,
236
+ "calibration_split": "train",
237
+ "eval_split": "validation",
238
+ "data_keys": {
239
+ "question": "question",
240
+ "context": "context"
241
+ },
242
+ "ref_keys": [
243
+ "answers"
244
+ ],
245
+ "max_seq_length": 128
246
+ },
247
+ "quantization_approach": "static",
248
+ "operators_to_quantize": [
249
+ "Add"
250
+ ],
251
+ "node_exclusion": [],
252
+ "aware_training": false,
253
+ "per_channel": false,
254
+ "calibration": {
255
+ "method": "minmax",
256
+ "num_calibration_samples": 20,
257
+ "calibration_histogram_percentile": null,
258
+ "calibration_moving_average": null,
259
+ "calibration_moving_average_constant": null
260
+ },
261
+ "framework": "onnxruntime",
262
+ "framework_args": {
263
+ "opset": 15,
264
+ "optimization_level": 1
265
+ },
266
+ "hardware": "Architecture: x86_64\nCPU op-mode(s): 32-bit, 64-bit\nByte Order: Little Endian\nAddress sizes: 46 bits physical, 48 bits virtual\nCPU(s): 20\nOn-line CPU(s) list: 0-19\nThread(s) per core: 1\nCore(s) per socket: 14\nSocket(s): 1\nNUMA node(s): 1\nVendor ID: GenuineIntel\nCPU family: 6\nModel: 154\nModel name: 12th Gen Intel(R) Core(TM) i7-1280P\nStepping: 3\nCPU MHz: 2000.000\nCPU max MHz: 4800.0000\nCPU min MHz: 400.0000\nBogoMIPS: 3993.60\nVirtualization: VT-x\nL1d cache: 336 KiB\nL1i cache: 224 KiB\nL2 cache: 8.8 MiB\nNUMA node0 CPU(s): 0-19\nVulnerability Itlb multihit: Not affected\nVulnerability L1tf: Not affected\nVulnerability Mds: Not affected\nVulnerability Meltdown: Not affected\nVulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp\nVulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2: Mitigation; Enhanced IBRS, IBPB conditional, RSB filling\nVulnerability Srbds: Not affected\nVulnerability Tsx async abort: Not affected\nFlags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l2 invpcid_single cdp_l2 ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdt_a rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req umip pku ospke waitpkg gfni vaes vpclmulqdq tme rdpid movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr flush_l1d arch_capabilities\n",
267
+ "versions": {
268
+ "transformers": "4.19.2",
269
+ "optimum": "1.2.1.dev0",
270
+ "optimum_hash": "f5a741330bef093b6ffe0023455dff722e4515e8"
271
+ },
272
+ "evaluation": {
273
+ "time": [
274
+ {
275
+ "batch_size": 8,
276
+ "input_length": 128,
277
+ "baseline": {
278
+ "nb_forwards": 9,
279
+ "throughput": 4.5,
280
+ "latency_mean": 241.8384202222222,
281
+ "latency_std": 14.81797115866389,
282
+ "latency_50": 231.185974,
283
+ "latency_90": 259.980655,
284
+ "latency_95": 260.193991,
285
+ "latency_99": 260.3646598,
286
+ "latency_999": 260.40306028
287
+ },
288
+ "optimized": {
289
+ "nb_forwards": 8,
290
+ "throughput": 4.0,
291
+ "latency_mean": 257.84814475,
292
+ "latency_std": 18.419393322852148,
293
+ "latency_50": 253.0442845,
294
+ "latency_90": 283.717168,
295
+ "latency_95": 288.036693,
296
+ "latency_99": 291.492313,
297
+ "latency_999": 292.2698275
298
+ }
299
+ },
300
+ {
301
+ "batch_size": 4,
302
+ "input_length": 128,
303
+ "baseline": {
304
+ "nb_forwards": 16,
305
+ "throughput": 8.0,
306
+ "latency_mean": 126.7981820625,
307
+ "latency_std": 7.175572633772116,
308
+ "latency_50": 124.789724,
309
+ "latency_90": 136.7108285,
310
+ "latency_95": 139.93856475,
311
+ "latency_99": 141.57276255000002,
312
+ "latency_999": 141.940457055
313
+ },
314
+ "optimized": {
315
+ "nb_forwards": 17,
316
+ "throughput": 8.5,
317
+ "latency_mean": 118.20609435294116,
318
+ "latency_std": 8.998285032579696,
319
+ "latency_50": 113.082448,
320
+ "latency_90": 131.37506059999998,
321
+ "latency_95": 132.626902,
322
+ "latency_99": 133.71953240000002,
323
+ "latency_999": 133.96537424
324
+ }
325
+ }
326
+ ],
327
+ "others": {
328
+ "baseline": {
329
+ "exact_match": 90.0,
330
+ "f1": 90.0
331
+ },
332
+ "optimized": {
333
+ "exact_match": 70.0,
334
+ "f1": 76.33333333333333
335
+ }
336
+ }
337
+ },
338
+ "metrics": [
339
+ "squad"
340
+ ],
341
+ "model_type": "distilbert"
342
+ },
343
+ {
344
+ "model_name_or_path": "distilbert-base-uncased-distilled-squad",
345
+ "task": "question-answering",
346
+ "dataset": {
347
+ "path": "squad",
348
+ "name": null,
349
+ "calibration_split": "train",
350
+ "eval_split": "validation",
351
+ "data_keys": {
352
+ "question": "question",
353
+ "context": "context"
354
+ },
355
+ "ref_keys": [
356
+ "answers"
357
+ ],
358
+ "max_seq_length": 128
359
+ },
360
+ "quantization_approach": "dynamic",
361
+ "operators_to_quantize": [
362
+ "Add"
363
+ ],
364
+ "node_exclusion": [],
365
+ "aware_training": false,
366
+ "per_channel": false,
367
+ "calibration": {
368
+ "method": "minmax",
369
+ "num_calibration_samples": 20,
370
+ "calibration_histogram_percentile": null,
371
+ "calibration_moving_average": null,
372
+ "calibration_moving_average_constant": null
373
+ },
374
+ "framework": "onnxruntime",
375
+ "framework_args": {
376
+ "opset": 15,
377
+ "optimization_level": 1
378
+ },
379
+ "hardware": "Architecture: x86_64\nCPU op-mode(s): 32-bit, 64-bit\nByte Order: Little Endian\nAddress sizes: 46 bits physical, 48 bits virtual\nCPU(s): 20\nOn-line CPU(s) list: 0-19\nThread(s) per core: 1\nCore(s) per socket: 14\nSocket(s): 1\nNUMA node(s): 1\nVendor ID: GenuineIntel\nCPU family: 6\nModel: 154\nModel name: 12th Gen Intel(R) Core(TM) i7-1280P\nStepping: 3\nCPU MHz: 2000.000\nCPU max MHz: 4800.0000\nCPU min MHz: 400.0000\nBogoMIPS: 3993.60\nVirtualization: VT-x\nL1d cache: 336 KiB\nL1i cache: 224 KiB\nL2 cache: 8.8 MiB\nNUMA node0 CPU(s): 0-19\nVulnerability Itlb multihit: Not affected\nVulnerability L1tf: Not affected\nVulnerability Mds: Not affected\nVulnerability Meltdown: Not affected\nVulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl and seccomp\nVulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization\nVulnerability Spectre v2: Mitigation; Enhanced IBRS, IBPB conditional, RSB filling\nVulnerability Srbds: Not affected\nVulnerability Tsx async abort: Not affected\nFlags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l2 invpcid_single cdp_l2 ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid rdt_a rdseed adx smap clflushopt clwb intel_pt sha_ni xsaveopt xsavec xgetbv1 xsaves split_lock_detect avx_vnni dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp hwp_pkg_req umip pku ospke waitpkg gfni vaes vpclmulqdq tme rdpid movdiri movdir64b fsrm md_clear serialize pconfig arch_lbr flush_l1d arch_capabilities\n",
380
+ "versions": {
381
+ "transformers": "4.19.2",
382
+ "optimum": "1.2.1.dev0",
383
+ "optimum_hash": "f5a741330bef093b6ffe0023455dff722e4515e8"
384
+ },
385
+ "evaluation": {
386
+ "time": [
387
+ {
388
+ "batch_size": 4,
389
+ "input_length": 128,
390
+ "baseline": {
391
+ "nb_forwards": 16,
392
+ "throughput": 8.0,
393
+ "latency_mean": 130.991486375,
394
+ "latency_std": 9.795692761346684,
395
+ "latency_50": 125.480088,
396
+ "latency_90": 146.3329905,
397
+ "latency_95": 148.674304,
398
+ "latency_99": 149.0314936,
399
+ "latency_999": 149.11186125999998
400
+ },
401
+ "optimized": {
402
+ "nb_forwards": 22,
403
+ "throughput": 11.0,
404
+ "latency_mean": 93.42974659090909,
405
+ "latency_std": 4.217314843877448,
406
+ "latency_50": 91.522541,
407
+ "latency_90": 99.6292669,
408
+ "latency_95": 100.35914185,
409
+ "latency_99": 102.45238115,
410
+ "latency_999": 102.944591015
411
+ }
412
+ },
413
+ {
414
+ "batch_size": 8,
415
+ "input_length": 128,
416
+ "baseline": {
417
+ "nb_forwards": 9,
418
+ "throughput": 4.5,
419
+ "latency_mean": 234.6135817777778,
420
+ "latency_std": 13.101986978479236,
421
+ "latency_50": 230.497678,
422
+ "latency_90": 251.6402062,
423
+ "latency_95": 252.2600866,
424
+ "latency_99": 252.75599092,
425
+ "latency_999": 252.86756939199998
426
+ },
427
+ "optimized": {
428
+ "nb_forwards": 11,
429
+ "throughput": 5.5,
430
+ "latency_mean": 191.77790254545454,
431
+ "latency_std": 3.7816204369571365,
432
+ "latency_50": 190.165539,
433
+ "latency_90": 198.02209,
434
+ "latency_95": 199.1632445,
435
+ "latency_99": 200.0761681,
436
+ "latency_999": 200.28157591
437
+ }
438
+ }
439
+ ],
440
+ "others": {
441
+ "baseline": {
442
+ "exact_match": 90.0,
443
+ "f1": 90.0
444
+ },
445
+ "optimized": {
446
+ "exact_match": 90.0,
447
+ "f1": 90.0
448
+ }
449
+ }
450
+ },
451
+ "metrics": [
452
+ "squad"
453
+ ],
454
+ "model_type": "distilbert"
455
+ }
456
+ ]
tensorboard/1652945902.9702048/events.out.tfevents.1652945902.huggingface.8.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:388a13d94c408ee5a93b9417a9295f3c7f8a253ddad8d008575007e065dfabaa
3
+ size 962
tensorboard/1652945902.971785/events.out.tfevents.1652945902.huggingface.8.2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a6b9fd3e097f3b7b8d95e35f11262747dff22279db0a2d97676967533e4dd17
3
+ size 952
tensorboard/1652945902.9734063/events.out.tfevents.1652945902.huggingface.8.3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95608d9e55688fe7c8369a0cf92751af284127df8d4835604fd3440173eb0d21
3
+ size 961
tensorboard/1652945902.9749742/events.out.tfevents.1652945902.huggingface.8.4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:deb33e8e84176640697894bbaed850e7c26b3b1d8f259ec0cc8ca6e923c7a941
3
+ size 951
tensorboard/events.out.tfevents.1652945902.huggingface.8.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ad08f531c5dbedf2dfa36a0c59d623a049228dc315b591ba52a9ceba31dab67
3
+ size 40